Appropriate problems for deep learning

[richardfichera]

Daniel,

I am at ~5hr23min of part 1. Earlier I believe you mentioned that problems for which there were existing solutions might not be good candidates for a TensorFlow solution. I do understand that I am slowly learning the nuts and bolts of TF as we work through the linear regression model, but since linear regression is a solution for which there are thousands of applications, would this really be a problem you would solve with TF?

BTW, I echo other comments I have read - fabulous course and great instructor. I am 70+, looking for something interesting to learn after Python (last serious programming was ~40 yrs ago n FORTRAN and assembly language, so this is a brave new world for me:)

[gauravreddy08]

Tensorflow works the best on non-tabulated datasets like images, speech and so on..
While on other hand, Regression you are talking about is a tabulated data.

In practice, people usually use "Machine Learning" (not Deep Learning) for regression problems.

Check Daniel's ML Course for learning ML using Scikit Learn which work best for regression problems.

[richardfichera]

Thank you. From: Gaurav Reddy ***@***.***> Reply-To: mrdbourke/tensorflow-deep-learning ***@***.***> Date: Sunday, May 23, 2021 at 3:57 AM To: mrdbourke/tensorflow-deep-learning ***@***.***> Cc: richardfichera ***@***.***>, Author ***@***.***> Subject: Re: [mrdbourke/tensorflow-deep-learning] Appropriate problems for deep learning (#66) Tensorflow works the best on non-tabulated datasets like images, speech and so on.. While on other hand, Regression you are talking about is a tabulated data. In practice, people usually use "Machine Learning" (not Deep Learning) for regression problems. Check Daniel's ML Course<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Facademy.zerotomastery.io%2Fp%2Fcomplete-machine-learning-and-data-science-bootcamp-zero-to-mastery&data=04%7C01%7C%7C2925299f6614401afcac08d91dc06038%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637573534345878247%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=rNmJi5qT8zY1O5%2BVdXToAqQXxv%2BZk%2BAk81l%2B%2F0G%2FkH4%3D&reserved=0> for learning ML using Scikit Learn which work best for regression problems. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fmrdbourke%2Ftensorflow-deep-learning%2Fdiscussions%2F66%23discussioncomment-773256&data=04%7C01%7C%7C2925299f6614401afcac08d91dc06038%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637573534345888202%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=d5sEQa5d%2BSjJ7Olc7gcpJkIeoS1zBFMQ1Y9RsKcMYOs%3D&reserved=0>, or unsubscribe<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FAN5FGPUAH6H53VIWFGQZO7DTPCYNRANCNFSM45KRUKYA&data=04%7C01%7C%7C2925299f6614401afcac08d91dc06038%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637573534345888202%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=ytOP86t3KG5eSDm9oI7pklJs6lpPqZX2%2FXh27sXXDpo%3D&reserved=0>.

[mrdbourke]

Hey Richard,

I am at ~5hr23min of part 1. Earlier I believe you mentioned that problems for which there were existing solutions might not be good candidates for a TensorFlow solution. I do understand that I am slowly learning the nuts and bolts of TF as we work through the linear regression model, but since linear regression is a solution for which there are thousands of applications, would this really be a problem you would solve with TF?

You're right, we don't necessarily have to use TF for linear regression problems.

That example is mainly to show that TF has a wide range of capabilities, from linear regression right up to building state of the art deep learning models.

It's really a production-ready numerical computing library. Almost anything you can think of numerically can be done with TensorFlow.

BTW, I echo other comments I have read - fabulous course and great instructor. I am 70+, looking for something interesting to learn after Python (last serious programming was ~40 yrs ago n FORTRAN and assembly language, so this is a brave new world for me:)

Thank you so much for the kind words! I really appreciate it.